[Bug 45045] New: Vampire: The Masquerade Bloodlines, GL_OUT_OF_MEMORY, worked in wine 1.8

https://bugs.winehq.org/show_bug.cgi?id=45045

Bug ID: 45045
           Summary: Vampire: The Masquerade Bloodlines,  GL_OUT_OF_MEMORY,
                    worked in wine 1.8
           Product: Wine
           Version: 3.6
          Hardware: x86-64
                OS: Linux
            Status: UNCONFIRMED
          Severity: normal
          Priority: P2
         Component: -unknown
          Assignee: [e-mail]
          Reporter: [e-mail]
      Distribution: ---

Created attachment 61172
  --> https://bugs.winehq.org/attachment.cgi?id=61172
Console outputs and backtraces for different videomemorysize settings.

Hello everybody,

I would like to report a crash of Vampire: The Masquerade Bloodlines (VTMB)
upon or shortly after loading of a saved game just after a fresh start of the
main game and/or area transition. It worked fine with wine 1.6 (debian jessie)
and the wine 1.8 from debian 9 (stretch, stable). It crashes with wine 3.0 and
wine 3.6 as detailed below.

The crash might be related to https://bugs.winehq.org/show_bug.cgi?id=42732 and
it has superficial similarities to
https://bugs.winehq.org/show_bug.cgi?id=24701   but because it worked for me
without problems in the earlier wine versions mentioned above I believe 24701
can be ruled out and I would like to file a new bug report explicitely. Please
feel free to eventually mark as duplicate as appropriate. I have some hope that
it might help people with the same problem anyway if they can find it.

The VTMB in question is the version distributed by GOG including the Unoffical
Patch version 9.7 by wesp5 as distributed by GOG (plus some additional
textures). The OS is debian 9 (stretch, stable) on a Haswell with IGP HD4400,
before upgrading I ran it on debian jessie (8, old-stable) on the same
hardware.

Again: No problems were observed with wine 1.6 and wine 1.8.

With wine 3.0 and wine 3.6 (I will only report/attach wine 3.6 details below) I
now observe a memory usage related crash upon (usually) loading of a saved game
and/or area transition depending on "winetricks videomemorysize=". The game was
newly started each time, i.e. it does not crash after hours of playing but
basically immediately.

I see on the console 
0038:err:d3d:wined3d_debug_callback 0x1ba070: "GL_OUT_OF_MEMORY in ...
before it crashes. I attached the console outputs and backtraces for various 
videomemorysize settings, i.e. default, 512, 1024, 2048. They do not make a
noticeable difference.

I then set LARGE_ADDRESS_AWARE as suggested in 42732 on vampire.exe. With this
modified binary I could not observe any problems when loading saves or doing
area transitions, but I tested it only for a few minutes. Still, this is an
obvious change.

So, in summary some change in wine after wine 1.8 might have broken that game
by somehow changing the memory layout.

Please let me know if I should gather any additional information.

Best Regards

Christof

P.S.: I tried to submit the same bug previously but it somehow never showed up
in the bug list. Not sure what happened.

[Bug 45045] Vampire: The Masquerade Bloodlines, GL_OUT_OF_MEMORY, worked in wine 1.8

https://bugs.winehq.org/show_bug.cgi?id=45045

--- Comment #1 from [e-mail] ---
Correction:

transition depending on "winetricks videomemorysize="

should read

transition not depending on "winetricks videomemorysize="

[Bug 45045] Vampire: The Masquerade Bloodlines, GL_OUT_OF_MEMORY, worked in wine 1.8

https://bugs.winehq.org/show_bug.cgi?id=45045

joaopa  changed:

What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |[e-mail]

--- Comment #2 from joaopa  ---
To have a good chance to see the bug fixed, you should do a regression test:
https://wiki.winehq.org/Regression_Testing

And the usual question: is there a freely downloadable demo showing the
problem.

[Bug 45045] Vampire: The Masquerade Bloodlines, GL_OUT_OF_MEMORY, worked in wine 1.8

https://bugs.winehq.org/show_bug.cgi?id=45045

Austin English  changed:

What    |Removed                     |Added
----------------------------------------------------------------------------
           Keywords|                            |regression

[Bug 45045] Vampire: The Masquerade Bloodlines, GL_OUT_OF_MEMORY, worked in wine 1.8

https://bugs.winehq.org/show_bug.cgi?id=45045

tokktokk  changed:

What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |[e-mail]

[Bug 45045] Vampire: The Masquerade Bloodlines, GL_OUT_OF_MEMORY, worked in wine 1.8

https://bugs.winehq.org/show_bug.cgi?id=45045

--- Comment #3 from [e-mail] ---
Hello everybody,

no, there (unfortunately) is no free reproducer.

I am aware that this might make it unfixable without detailed regression test
but hoped that full regression testing could be avoided somehow. I once built
wine from source on a workstation and hunting down and installing the
dependencies took half a day or so and I believe compiling was another hour or
two. I fear that this is not good on a laptop, especially as the SSD is low on
free space and write endurance.

However, I notice that winehq has debian packages as far back as 2.0 (as far as
I could see all of them in a stretch version). Would it be helpful to try these
package releases one after the other as a first step ? Bisecting on the commit
granularity could be done later.

I would have to move that to a VM on a "real" computer anyway. So the
preparation will take quite some time I am afraid.

In any case thank you for your interest and effort !

Best Regards

Christof

[Bug 45045] Vampire: The Masquerade Bloodlines, GL_OUT_OF_MEMORY, worked in wine 1.8

https://bugs.winehq.org/show_bug.cgi?id=45045

--- Comment #4 from joaopa  ---
So, the bug does not occur with 
https://www.fileplanet.com/46322/download/Vampire:-The-Masquerade-Redemption...

?

[Bug 45045] Vampire: The Masquerade Bloodlines, GL_OUT_OF_MEMORY, worked in wine 1.8

https://bugs.winehq.org/show_bug.cgi?id=45045

--- Comment #5 from [e-mail] ---
Created attachment 61186
  --> https://bugs.winehq.org/attachment.cgi?id=61186
Console log and backtrace with 1.8, vidoememorysize=default

[Bug 45045] Vampire: The Masquerade Bloodlines, GL_OUT_OF_MEMORY, worked in wine 1.8

https://bugs.winehq.org/show_bug.cgi?id=45045

--- Comment #6 from [e-mail] ---
Hello everybody,

I discovered that the newest saves from the 3rd playthrough are from February
4th. So they are quite old. I reinstalled wine 1.8 then to make sure everything
was as I assumed it was and it turns out that it now also runs out of memory
with wine 1.8 ...

err:d3d:wined3d_debug_callback 0x210b68: "GL_OUT_OF_MEMORY in
glMapBufferRange(map failed)".

On the other hand it must have worked on some point as I have late game saves
from two complete previous playthroughs.

It might be that I only played it with wine 1.6 from jessie. On the other hand 
I have a lvm.conf.dpkg-new dated February 1st which would hint that it is that
date on which I upgraded to stretch. In that case the February 4th saves should
be with wine 1.8.
I uploaded a console log and backtrace with wine 1.8.

I am very confused now and I would like to apologize for apparently not testing
thoroughly enough.

If I find a way to install wine 1.6 from jessie on the current OS I will try
that. Eventually this is more OS/i915 driver related than it is wine related
and this should be ruled out first of course. Testing in a VM might yield some
clues.

If you have any more input on how to procede I would appreciate it. I will
report back in any case.

Please note that Redemption is a completely different game from a different
company with a different game engine, just using the same game license.
Bloodlines uses an early Source Engine (-> Half Life 2). So trying that demo
does not make sense.

Best Regards

Christof

[Bug 45045] Vampire: The Masquerade Bloodlines, GL_OUT_OF_MEMORY, worked in wine 1.8

https://bugs.winehq.org/show_bug.cgi?id=45045

--- Comment #7 from [e-mail] ---
Hello again,

now I am back on wine 3.6 and tried to load different saves. As it turns out
one  indoor save can be loaded and it is playable (inside a single area, you
cannot easily leave because it is end game). The save I used for the original
report, which is the newest save I have and also indoors, still does crash in
the same way. I alsoe found one where the game crashes reliably on area
transition from indoors to outdoors. I assume this depends on the order/amount
of game data to be loaded into memory. The one which crashes on area transition
gives in the console
0059:err:d3d:wined3d_resource_allocate_sysmem Failed to allocate system memory.
which I believe is a new one.

I will think about how to come to a conclusion on this tomorrow.

Best Regards

Christof

[Bug 45045] Vampire: The Masquerade Bloodlines, GL_OUT_OF_MEMORY, worked in wine 1.8

https://bugs.winehq.org/show_bug.cgi?id=45045

--- Comment #8 from [e-mail] ---
Created attachment 61190
  --> https://bugs.winehq.org/attachment.cgi?id=61190
Backtrace with wine 1.6 from jessie running on stretch

[Bug 45045] Vampire: The Masquerade Bloodlines, GL_OUT_OF_MEMORY, worked in wine 1.8

https://bugs.winehq.org/show_bug.cgi?id=45045

--- Comment #9 from [e-mail] ---
Hello everyobody,

I have to admit defeat with this at least for the moment.
> From various file creation dates on the system I am sure that the upgrade from

jessie to stetch happened on February 1st. And then I played it on February 4th
for several hours producing several saves. That must have been with wine 1.8.
So it must have worked.

I now installed the jessie wine 1.6 (wine32) packages on top of the stretch
system. And: even with wine 1.6 it crashes at the same points.
A backtrace (top is  0 0x5631ff0c in i965_dri.so) is attached.

So I have to admit that I am no longer able to reproduce the condition in which
it worked to begin with. This probably indicates that wine is not the problem
but some other factor. I am not sure if I will be able to identify it.

Because of this you may close this bug report either as invalid or, if you
prefer, leave it open and I will report if I happen to be able either to reach
the working state again and/or what happens in a VM (different graphics driver
etc.).

Thank you very much for your interest and my apologies again for not testing
thoroughly enough right at the start but just assuming it was the wine version.

Best Regards

Christof

[Bug 45045] Vampire: The Masquerade Bloodlines, GL_OUT_OF_MEMORY, worked in wine 1.8

https://bugs.winehq.org/show_bug.cgi?id=45045

Matteo Bruni  changed:

What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |RESOLVED
         Resolution|---                         |INVALID

--- Comment #10 from Matteo Bruni  ---
I'm resolving this as INVALID for now, but feel free to reopen if you find new
info (e.g. you figure out it's a Mesa regression, or whatever happens.)

[Bug 45045] Vampire: The Masquerade Bloodlines, GL_OUT_OF_MEMORY, worked in wine 1.8

https://bugs.winehq.org/show_bug.cgi?id=45045

--- Comment #11 from [e-mail] ---
Hello,

OK, great !

One small bit of info: The modified exe with LARGE_ADDRESS_AWARE set continous
to just work. The modified exe uses about VSZ 2.2 GByte/ RSS 680 MByte after
loading a safe. The original exe at the time it crashes is using VSZ 3.1 GByte/
RSS 220 MByte.

Best Regards

Christof

[Bug 45045] Vampire: The Masquerade Bloodlines, GL_OUT_OF_MEMORY, worked in wine 1.8

https://bugs.winehq.org/show_bug.cgi?id=45045

André H.  changed:

What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|RESOLVED                    |CLOSED
                 CC|                            |[e-mail]

--- Comment #12 from André H.  ---
closing invalid